3.4.2. Prediction Process Used by the CA–Markov Model

The CA–Markov module in IDRISI systematically integrates the Markov chain model with the CA model. The Multi-Criteria Evaluation module can be used for the prediction of land-use spatial patterns and the analysis of land transfer suitability. We simulated the land-use pattern with the following steps:


type reserves, biological species reserves, and natural heritage reserves. Influencing factors enhance or detract from the suitability of a location for the objective. All influencing factor images must be measured on the same scale. In this study, the scale 0–255 is used where 0 is not at all suitable and 255 is perfectly suitable. Then the weights of the influencing factors are produced follows the logic under the Analytical Hierarchy Process (AHP). That is, we should provide a series of pairwise comparisons of the relative importance of factors to the suitability of pixels for the activity being evaluated. (Table 1). Second, the constraints were transformed into a binary image. That is, areas with a construction land spatial distribution probability of zero were assigned a value of 0. The factors were transformed into standard data in the range of 0–255 using the Fuzzy module, and the fuzzy set membership function was applied to evaluate the effects of factors on suitability. Finally, the MCE module was used to generate a construction land transition suitability image.

(3) Finally, the land-use spatial pattern in 2030 was predicted. We selected default 5 × 5 filters as the neighborhood scale, which is suitable for most simulation processes to ensure accuracy and also avoid running the simulation too far into the future. Then the number of iterations was set to 8. Land-use images from 2000 and 2010 were regarded as earlier and later images, respectively. We input the land-use transfer matrix and Transition suitability image collection so that we obtained a predicted land-use image for 2018. Then, the Crosstab module was applied to compare this with the actual image of land use in 2018 to get a kappa coefficient of 0.9243, indicating that the simulation effect was good. Therefore, the actual land-use image in 2018 could be served as the early image to simulate the land-use image in 2030 using the method above.


**Table 1.** Driving factors identified in the suitability analysis of construction land.

*3.5. Low-Carbon Optimization Model of Land Use*
